Research Questions

1. Are the effects of climate change (specifically air pollution) evident in epidemiological prevalence data or mortality data?

2. How do climate change ripple effects (measured via infant and maternal mortality) show differently in developing (i.e. more agriculture-based) versus developed countries?

3. How has the impact of natural disasters changed over time? How many people have been affected by these climate disasters?

Investigation on Air Quality through Leaflets

To visualize our data, we chose to create a series of choropleths that display the level of exposure to particulate matter and the number of years lost due to this exposure in each country around the world. The following two choropleths show PM2.5 exposure in 1990 and 2017, respectively. We chose to create leaflets to allow for interactivity of hovering over or selecting a particular country.

To help with the understanding of this measure, we will provide a few examples in context. When a user selects the country Brazil by clicking on it on the leaflet showing the data for 1990, a value of 15.402 µg/m³ will be displayed as the PM2.5 measure. This means that the mean annual outdoor PM2.5 concentration is 15.402 µg/m³, and this value is also weighted, or standardized, by population. As another example, Saudi Arabia displays a value of 92.368 µg/m³ on the leaflet showing data for 2017. Based on the PM2.5 guideline provided by the World Health Organization, Brazil in 1990 is considered to be in a moderate range of PM2.5, while Saudi Arabia is in a highly dangerous range. Despite such great differences in the PM2.5 level in different countries, these two choropleths show that the air quality in general has improved over the years from 1990 to 2017. This change is most noticeable in many countries in the Eastern Europe, Latin America, and Southern Africa, while there are several countries undergoing a similar change in parts of Asia.

These leaflets are effectively mapping the PM2.5 level on a global scale and letting us compare the PM2.5 level between multiple countries. Though we can see some changes undergone by some countries, this change is not dramatically apparent at first sight. Therefore, we created an additional leaflet that shows the change in PM2.5 level from 1990 to 2017. The value indicates the change from 1990 to 2017 (i.e. value_1990 - value_2017), so a positive value indicates a decrease in PM2.5 level, while a negative value indicates an increase. For example, India experienced an increase in PM2.5 level of 8.558 µg/m³. As we expect, now more clearly, we see a positive change in air quality in general around the world.

Investigation on Mortality (DALY) through Leaflets

Similar steps were taken for the next set of choropleths focusing on the disability-adjusted life years in 1990 and 2017 around the world. Just to interpret a few observations in context, we see that people in Brazil in 1990 lost 5.77 years of potential life due to exposure to ambient PM2.5 per 1000 inhabitants, and those in Russia in 2017 lost 15.681 years. Comparing these two choropleths, we see in general a decrease in the DALY score around the world, and this trend is most apparent in countries in Eastern Europe and Africa.

However, a similar concern arises as before: though we are effectively showing the distribution of DALY around the world for each year, the change is not necessarily clear at first sight due to a relatively small range of values. Therefore, similar steps were taken to show the change in DALY from 1990 to 2017. Note that the change was calculated by subtracting the value in 2017 from the value in 1990 (i.e. change = DALY_1990 - DALY_2017). Therefore, a positive value on this map indicates a decrease in DALY, while a negative value indicates an increase in DALY. For example, Spain has undergone a positive change (decrease) in DALY. Consistent with the two maps created above, we see that in general the countries experience a decrease in DALY.

Now, taking a few steps back to the original question we are trying to answer, notice that the two maps showing the change in PM2.5 and DALY look quite similar at a first glance. There are countries that show an opposing trend, where there is an increase in PM2.5 exposure and a decrease in DALY, and vice versa. This might be accounted by other factors that might not have been reflected in this study, such as noticeable improvements in medical technology that might have contributed to a decrease in DALY despite the worsened air quality. However, for the most part, the countries that have a positive change in PM2.5 also experience a positive change in DALY, so we can reasonably conclude that PM2.5, an environmental factor, is evident in and consistent with DALY, a mortality measure.

The relationship between PM2.5 and DALY through a scatterplot

The choropleths have sufficiently answered the question we wished to answer. However, a curious data scientist (like ourselves) might want to see more, perhaps what might be going on in between the two given years of 1990 and 2018, and whether the relationship between PM2.5 and DALY we have established earlier still holds true, if we were to take a look at a different time frame.

The most intuitive way to achieve this is probably to create plots that show the relationship between PM2.5 and DALY for a particular set of years, and without a Shiny app, we are almost certainly destined to make a multiple set of data visualizations, which can be exciting but also tedious and inefficient. In fact, this should be a supplementary resource, not a source of confusion or incompetence. Therefore, our solution to this is to retain the original vision to create a plot looking at how DALY changes with PM2.5, but produce these plots in a single gif file to show an animation of the change.

In the following plot, each colored dot represents each individual country available in our data. As we proceed from 1990 to 2017, we see that the points in general move towards the bottom-left corner from the top-right corner of the plot. A noticeable exception to this happens from 2014 to 2015 (try to observe this yourself!) where many data points move towards the right side of the plot, indicating that from 2014 to 2015 there was a relatively significant increase in PM2.5 exposure in many countries around the world. This may be something that might be worthwhile to take some time to research on; this analysis indeed revealed something that we would not have been able to see if we stopped our analysis after the first part.

What actually happened in between the years 1990 and 2017?

Now, with those in mind, an even more curious data scientist might want to see how either PM2.5 and DALY change for a particular country over a period of time that is not established through the choropleths. That is, a line graph or a similar sort where the x axis represents the year and the y axis represents either PM2.5 or DALY. This would be something that would be achieved most effectively and efficiently through shiny app’s interactivity: there are approximately 200 countries in the world (and in this data) and to display all of these countries at once would be unconventional. However, instead of making an extensive shiny app, we created a sample set of two line graphs that each show PM2.5 and DALY change over the years for a set of 10 randomly selected countries. We had done this because we did not want to put an equal or greater amount of coding efforts to something that is not the main tool to be used to achieve the original purpose, but still leave room for further investigation.

The following two interactive line graphs show the change in DM2.5 and DALY from 1990 to 2017. This is similar to the earlier change leaflets we had displayed but different in that these show all the values for a country in a given year and that these don’t account for the rest of approximately 190 countries. Therefore, we will not make a particular analysis on these visualizations, as our analysis will not account for the entirety of the data. Please consider these as simply a supplemental resource to make a similar analysis we made in order to answer our original question with an additional capability to observe the variables in a given year between 1990 and 2017.

Next Steps

As seen in our analysis using the choropleths, we had found that in general there seems to be a direct relationship between the change in air quality and mortality across countries. That is, an environmental factor can be evident in, and thus have directly impacted, the quality of life of the people. This being said, we wondered narrow down our focus to investigate whether an environmental factor could disproportionately affect countries around the world. Therefore, in the next section of this blog post, we will aim to answer the following question: “Are developing countries affected differently than developed countries by climate disasters?”

Infant and Maternal Mortality from 1980-2018

First, let’s consider how the distribution of countries’ infant and maternal mortality rates have changed over the years. Although more recent, even in the 1980s, we did not possess all the current medicine and technology we have today, especially medically related. The data on infant and maternal mortality that we used came from the OECD. Infant mortality was categorized by no minimum threshold of gestation period or birthweight, and recorded in terms of deaths per 1,000 live births. Maternal mortality was recorded in terms of deaths per 100,000 live births. From Figures 11 and 12 we can see that without looking at specific countries, there is a trend of decreasing infant and maternal mortality. In the 1980s, 23 countries were in our self-labeled “Developing” cluster (between standardized values of around -1 to 0), while in the 2010s, 31 countries are in the “Developing” cluster (note the range has increased from -1 to 1). What’s odd is that although we would expect a stark change in infant/maternal mortality values over time, we see more nuanced changes. The “Developing” cluster increases in size and range of values covered, while the “Middle” cluster is much more distinct and widespread than in the 1980s. We also still have one country (Mexico) that seemingly hasn’t changed in terms of standardized values. It is likely that from 1980s to the 2010s, the mean infant and maternal mortality rates greatly improved, but the distirbution and distinction between various country groupings (such as “Developed”, “Middle”, and “Developing”) has stayed fairly consistent. If infant mortality and maternal mortality all improved to the same standard, we would expect to see all countries in the “Developed” cluster. However, instead we see a larger “Developed” cluster, and blatant countries that have been left behind so to say, in terms of their infant and maternal mortality rates. All countries displayed reduction in infant/maternal mortality rates from the 1980s to 2010s. This can be better seen when looking at yearly values in the data set (use the link below to view the data used). However, from these clusters, we can draw our first conclusion, that not all countries have reached the same infant/maternal mortality rates. There is still a disparity in mortality rates for different countries.

One thing to note is that the data we used only had information on countries in North and South America, Australia, and parts of Europe. Many of the more predominant developing countries and agriculturally-based societies, such as in Africa and South/Southeast Asia did not have data recorded in the dataset.

Data Used:

Go to Health > Health Status > Maternal and infant mortality. Variables, years, and measures can be filtered out.

https://stats.oecd.org/Index.aspx?DatasetCode=HEALTH_STAT

## Parsed with column specification:
## cols(
##   COU = col_character(),
##   Country = col_character(),
##   decade = col_double(),
##   avginfmort = col_double(),
##   avmatmort = col_double(),
##   avginfmort_std = col_double(),
##   avmatmort_std = col_double(),
##   clust2 = col_double()
## )

What does Yearly Change in Infant and Maternal Mortality Look Like?

Let’s zoom in to look how infant mortality changes yearly for a pair of countries. Consider Switzerland, the second most developed nation as of 2020, according to life expectancy, economic competitiveness, infrastructure, GDP per capita (Developed Countries List 2020). Switzerland’s economy is based more in the service sector. Compare this to Mexico, which is less developed industrially, has more widespread poverty, and is a more agriculturally-based society. We know that Mexico would has worse infant and maternal mortality rates, compared to Switzerland (can be seen in original data tables). Specifically, Mexico started at 54.8 infant deaths per 1,000 live births in 1980 and decreased to 12.9 deaths per 1,000 live births in 2018. Switzerland started at 9.1 infant deaths per 1,000 live births in 1980 and decreased to 3.3 infant deaths per 1,000 live births in 2018. Thus, we can see that even in 1980, there was a great disparity in starting point and end point, however by 2018, the gap was greatly decreased. Now if we look at the yearly changes, we can understand how Mexico has much greater negative net changes in infant mortality. Note, I chose to focus the timeline on more recent years so that there is less impact from technological advancements in medicine. In fact, there is only one year (2017) where there is a positive net change in infant mortality. Switzerland on the other hand, while having negative net change in infant mortality between 2010 and 2018, has less frequent changes, and changes of lesser magnitude. In several years, the infant mortality rate stays the same. It also displays a positive net change in infant mortality in 2013. So it seems that in this case, the more agriculturally-based society has had a larger improvement in infant mortality rates than the classified Developed Country due to their largely varying starting points.

STILL NEED TO TYPE TEXT FOR:

-two country comparison of maternal mortality -two country comparison of natural disasters occurrence and mortality rates -general trends/associations for infant mortality, maternal mortality, and natural disasters -limitations/improvements/speculations

Data Used:

Natural Disaster Frequency Data by Country obtained from https://public.emdat.be/data

Note you must create an account and validate through email in order to view the datasets. Data is available from 1900-2020.

References Used:

https://worldpopulationreview.com/country-rankings/developed-countries https://worldpopulationreview.com/country-rankings/developing-countries

Is there an Association Between the Frequency of Climate Disasters and Infant/Maternal Mortality?

We can hypothesize that countries facing higher frequencies, and more intense, natural disasters risk more damages to their agricultural and resources systems, which can reduce access to proper food, jobs, etc. that might ensure lower infant and maternal mortality rates.

Frequency of Natural Disasters

First, we will test the hypothesis that the frequency of impactful natural disasters had increased over time. Take a look at a shiny app below.

• Compare the occurrences of different natural disasters. Are some natural disasters more common than others?
• What kind of general trend over time do you see?
• Select the disasters one by one. Did the frequency of some disasters increase more that the frequency of other disasters over time?

What do you see? First, we see that some natural disasters are more common than other disasters. Take the year 2011 as an example. There were 8 landslides in 2011 worldwide, but there were 84 floods. In year 2011, floods were 10 times more frequent than landslides, and this pattern holds generally true for other years as well. Floods and storms appear to be more frequent than droughts, extreme temperatures, landslides, and wildfires.

Next, we can observe that the frequency of some disasters increased more than the frequency of other disasters. Examine the frequency of Droughts from 1980 to 2020. There is a great deal of fluctuation in the annual global frequency of droughts, but ultimately the peaks in the graphs stay relatively stable. In 1999, there were 26 droughts, and so were in 2015. ( Keep in mind that we are only looking at the frequency per se, and not how it actually affected human societies. Save that thought for below) In contrast, examine the frequency of floods over time. We can see that floods became more and more frequent over time. There were only 76 floods in 1988, but 226 floods in 2006! In fact, the increasing trend is clearer for extreme temperatures, floods, and storms.

Most importantly, examining how the total occurrences of these 6 disasters, we can see that there is a general increasing trend. There is a noticeable increase in disaster frequency in 1990s, and it remains increased after 2000. We suspect that the increase may have been stalled after 2000 because countries became more equipped to handle natural disasters, suffered less damages from them. Such change would have rendered the disasters that were “impactful” in the past “impactless” (See the criteria in the shiny app above). Nevertheless, it is impossible to investigate how impactful these disasters were from this plot. For our second hypothesis, we turn to the interactive leaflet below.

Impacts of Natural Disasters by Region

Take a look at the shiny app below. We created an interactive leaflet to allow the user to select a year between 1980 and 2020 and the type of disaster they are interested in. Each region is colored according to the total number of people killed, injured, displaced, or affected by the natural disasters of chosen types.

• Start from 1980. Examine the choropleth every 4-5 years until you reach 2020. What kind of global trend do you see?
• Examine each of the disasters one by one. What trend, if any, do you see?

As we examine the plot from the earlier years to the more recent years, we noticed that the choropleth became more and more colorful. This means that more people were killed, injured, displaced, or affected by the aforementioned natural disasters in recent years.

There are a few caveats that you should keep in mind: This dataset does not record every single storm, drought, and flood that happened for the past few decades. Because it only records the natural disasters that affected humans beyond a certain extent, the trends shown here should be understood as a proxy, rather than a direct measurement of the frequency of natural disasters. For example, a storm may wreak havoc in developing countries that lack infrastructures to predict storms, prepare for the storms, and take necessary protective measures. However, an equally severe storm may barely affect some developed countries where houses and buildings are designed to withstand natural disasters, in which case the storm will not be recorded in this dataset.